So whether it's via robot or human spacewalk (though a manned operation is even more expensive obviously) the issue is the prohibitive expense of getting to the Hubble to fix it.
So 'getting to the Hubble' is pretty low in price, all things considered. Now the hard part: building something that can fix the Hubble. Humans would be good at making repairs, but the SpaceX Dragon capsule doesn't have an airlock, so that won't work easily. Repairs might be doable by a robot, but I suspect the easier option would be to simply attach new parts to it to perform the broken functionality.
If you can build the new piece for $40m, that's $100m to add another 20+ years to the lifespan of this incredible piece of equipment. Heck, sounds like something Bezos might do just for fun and as a test for his New Glenn rockets.
Can Dragon 2 do what most capsules do and just cycle the capsule interior into vacuum then open the door? (With everyone in suits first, obviously).
Neither did the Apollo moon landers, or the Gemini missions where NASA first tested spacewalks. Instead they just depressurized the whole cabin, making it into one big airlock. Presumably this would work for the Dragon as well.
I think it's instructive that they never tried that with any of the earlier Hubble repair missions.
At that point you’d just launch a new one. Earth’s gravity well assesses a steep tax.
The key thing to remember when it comes to space is energy/momentum. Right now, the Hubble Space Telescope is 24,000 lb of mass moving at 4.7 miles per second. To get it back down to earth, you need to somehow slow it down. First, slow it down enough to reach the atmosphere and then protect it as the atmo drag slowed it the rest of the way. I don't even think the Space Shuttle, were it still around, would be capable of doing that.
Remote repair using a robot is a whole other issue, and I don't think we have the tech developed to do this reliably any time soon.